I collect gold fingers in the heavens

Chapter 671 AI Drawing
In many fields, AI has been able to use thinking close to human beings to complete tasks more efficiently.

For example, in terms of machine translation, with the help of AI, the quality of translation has reached the level of usability in many scenarios, and computers and the Internet have become more and more intelligent.

The most outstanding thing in the world before is actually [Chaoqun Group], which launched [Smart Butler Xiaolan] for the world, which integrates the function of AI translation, and the translation effect is not worse than professional human translation. .

It's just that everyone paid attention to the intelligence of [Smart Housekeeper Xiaolan] in human-computer communication, while ignoring its outstanding performance in translation.

To put it bluntly, just using this function alone can create a thousand-level Internet company-if it is not blocked by overseas markets like [Super Group].

Um, Qiandu is going back more and more, and it is almost becoming a unit of measurement in the country, just like Nie Feng in "The Wind and Cloud World".

This function of [Smart Housekeeper Xiaolan] is only noticed by business personnel of some multinational companies and applied in foreign trade exchanges.

The effect is not bad.

【Smart Housekeeper Xiaolan】It is very sad for this group of people to be completely blocked in overseas markets, because there is no software with similar functions on the market!
Back to the AI.

I have to admit that AI has performed better and better in many fields, but in some fields, such as art, painting, writing and other fields that can reflect human original thinking, most people still think that AI cannot achieve human level , because humans have unlimited imagination, while AI is just a program.

It is true that in these aspects, the development of AI does not show the ability to surpass the human level, but its progress is also obvious to the naked eye.

For example, in the field of painting.

In recent years, the development of AI in graphics has been enough to surprise many people.

Many high-quality AI painting solutions have come out one after another. Even if the user can't draw at all and has no artistic talent at all, with a few graffiti, or even a simple description of one or two sentences or even a few words, it can be generated by AI. The beautiful paintings are like black technology.

You input a vocabulary of [Crane], and [AI Painting] can create many paintings that catch your eye, with various styles.

In fact, [AI Painting] does not have such a surprising effect in just one interview. Like other products, it has undergone countless development iterations.With the increase in computing power and the improvement of the algorithm, it has gradually cultivated to its present skill.

If you look closely, [AI Painting] may appear earlier than many people imagined.

The first computer appeared in the 60s, and in the 70s, an artist used a computer program to paint, but it was different from the current [AI painting] output digital works, the method at that time was to really control a machine It belongs to the hard core [AI painting].

In 2006, a computer painting product appeared, which can observe photos, extract block color information in photos, and use real painting materials such as paint, pastels or pencils to create, but in terms of intelligence See, this is very rudimentary.

The current [AI painting] refers more to the computer program for automatic drawing based on the deep learning model. The development of this drawing method is actually relatively late.

In 2012, an unprecedented experiment used 1000 million cat face pictures, and 1.6 CPUs were trained for 3 days, and the final model was exciting enough to generate a very blurry cat face.

In today's view, the training efficiency and output results of this model are not worth mentioning. But for the field of AI research at that time, this was a breakthrough attempt, which officially opened the [AI Painting] project supported by the deep learning model. "New" research direction.

Developed to the present, [AI painting] can be described by leaps and bounds, from the CLIP model based on massive Internet pictures that do not need to be labeled, to the upsurge of AI painting model grafting triggered by CLIP open source, and then found the Diffusion diffusion model as a better The image generation module finally uses the improved method of latent space dimensionality reduction to solve the problem of huge consumption of time and memory resources of the Diffusion model... All of this is dizzying, and the level of [AI painting] that has been stagnant for many years is rocket-like The speed rushed to the top.

Today's [AI painting] has developed to transform a novice work that seems to be full of mistakes into a picture with very good light, shadow and shape, and even enter the required keywords into it to produce a picture with a very high artistic level. The point of the painting.

But [AI Painting] is not perfect, there are still many loopholes in the field of creation, although some pictures are very beautiful at first glance.But the picture itself has no depth, and some are too stereotyped. After a long time, it will inevitably appear a little "aesthetic fatigue", that is to say, the artistic depth is not enough.

In addition, AI is currently not perfect in some recognition functions, and it is inevitable that there will be screen effects that "violate the common sense of characters".

Of course, this is another [AI painting], the [AI painting] launched by [Chaoqun Group] is not comparable to these vulgar products.

At least, with [Superior Group]'s mastery of artificial intelligence, there will be basically no recognition errors, let alone draw paintings that [violate human common sense].

The current [AI Painting] can only respond to some simple vocabulary and generate corresponding paintings.

However, the [AI Painting] launched by [Chaoqun Group] can recognize the content of an entire book with an intelligent program, thereby producing a coherent painting. And made into animation.

During this process, users can adjust the entire painting according to their own needs.

Take [Journey to the West] as an example.

After ingesting this book, the [AI Painting] program of [Chaoqun Group] can call out various elements generated from the application level.

For example, in terms of characters, the AI ​​program is based on the description to generate the images of the Monkey King, Bull Demon King, Tang Seng and other characters that best fit the book image. After the birth, the user can adjust the image through the tools provided by the program, or modify the program Say your own demands, and make minor or major revisions to these characters—or fully sissy them, or Krusu-like them. [Super Group]'s [AI Painting] program has this ability Support user requirements.

(End of this chapter)